You are asked to solve a business problem using NLP, and the right approach could range from a lightweight classifier to a domain-tuned transformer or a structured extraction pipeline. You need to decide how to frame the task, prepare text data, choose an architecture, and connect model outputs to a business workflow.
How would you approach Natural Language Processing (NLP) architectures for a business problem?